Microsatellite instability detection in cell-free DNA
Provided herein are methods for determining the microsatellite instability status of samples. In one aspect, the methods include quantifying a number of different repeat lengths present at each of a plurality of microsatellite loci from sequence information to generate a site score for each of the plurality of the microsatellite loci. The methods also include comparing the site score of a given microsatellite locus to a site specific trained threshold for the given microsatellite locus for each of the plurality of the microsatellite loci and calling the given microsatellite locus as being unstable when the site score of the given microsatellite locus exceeds the site specific trained threshold for the given microsatellite locus to generate a microsatellite instability score, which includes a number of unstable microsatellite loci from the plurality of the microsatellite loci.
1. A method of treatment, comprising:
(a) selecting a human subject having cancer characterized by microsatellite instability, comprising:
(i) generating a site score (SS) by quantifying a number of different repeat lengths present at each of a plurality of microsatellite loci from sequence information for each of the plurality of the microsatellite loci in each of a plurality of human subjects having cancer characterized by microsatellite instability (MSI),
wherein the sequence information is from a sample comprising cell-free deoxyribonucleic acids (cfDNA) molecules obtained from the human subject,
wherein the site scores of the plurality of the microsatellite loci comprise likelihood scores that comprise probabilistic log likelihood-based scores for an individual microsatellite locus determined using at least a first parameter comprising observed allele frequencies of the different repeat lengths and at least a second parameter comprising noise in the sequence information;
(ii) comparing the SS of a given microsatellite locus to a site specific trained threshold for the given microsatellite locus in each of the plurality of human subjects, wherein the site specific trained threshold is the maximum value of a site score for a microsatellite locus to be classified as stable, for each of the plurality of the microsatellite loci;
(iii) calling the given microsatellite locus as being unstable if the SS of the given microsatellite locus exceeds the site specific trained threshold for the given microsatellite locus in each of the plurality of human subjects, to further generate a microsatellite instability (MI) score comprising at least one unstable microsatellite loci from the plurality of the microsatellite loci;
(iv) classifying microsatellite instability status (MSI-S) of the sample as being unstable in each of the plurality of human subjects, if the MI score exceeds a population trained threshold for the population of microsatellite loci in the sample to identify an unstable sample; and
(b) administering at least one immunotherapy to at least one of the plurality of human subjects whose MSI status is classified as unstable in step (iv) to the human subject, wherein the at least one immunotherapy comprises at least one of:
an immune checkpoint molecule,
an antibody specific for an antigen selected from the group consisting of: PD-1, PD-2, PD-L1, PD-L2, CTLA-4, OX40, B7.1, B7He, LAG3, CD137, KIR, CCR5, CD27, CD40, and CD47,
proinflammatory cytokine selected from the group consisting of: IL-1β, IL-6, and TNF-α, and
activated T-cells, thereby treating the cancer in the human subject.
2. The method of claim 1 , wherein the site score of score of a given microsatellite locus comprises an AIC-based site score that tests for a presence of somatic indels at that microsatellite locus for an alternate hypothesis that the given microsatellite locus is unstable.
3. The method of claim 2 , wherein the site score of the given microsatellite locus comprises Akaike Information Criterion (AIC)-based sites.
4. The method of claim 3 , wherein a given AIC-based site score is calculated using the formula of:
AIC
=
k
-
log
-
likelihood
,
where k is the number of parameters used in a model for generating the site scores.
5. The method of claim 4 , comprising calculating a null hypothesis score of the model using the formula of:
AIC
0
=
k
-
log
(
P
r
(
obs
|
β
,
γ
)
)
,
where AIC 0 is the null hypothesis, k is the number of parameters used in the model, Pris probability, obs comprises repeat lengths of observed sequencing reads covering the given microsatellite locus, β is at least one strand specific error parameter, and γ is at least one random error parameter.
6. The method of claim 5 , comprising calculating an alternate hypothesis score of the model using the formula of:
AIC
min
=
min
α
(
k
-
log
(
P
r
(
obs
|
β
,
γ
,
α
)
)
,
where AIC min is the alternate hypothesis, mina an effect of minimizing over all values of α, k is the number of parameters used in the model, Pr is probability, obs comprises repeat lengths of observed sequencing reads covering the given microsatellite locus, β is at least one strand specific error parameter, γ is at least one random error parameter, and α is at least one allele frequency.
7. The method of claim 6 , comprising detecting the site score using the formula of:
Δ
AIC
=
AIC
0
-
AIC
min
.
8. The method of claim 6 , wherein γ comprises:
a rate of read-level errors where a microsatellite length observed within a sequencing read is one repeat unit longer than an expected microsatellite length for a strand of an originating cfDNA molecule; and/or
a rate of read-level errors where a microsatellite length observed within a sequencing read is one repeat unit shorter than an expected microsatellite length for a strand of an originating cfDNA molecule and/or wherein β comprises:
a rate of strand-level errors where an expected microsatellite length of a sense strand is one repeat unit longer than an expected microsatellite length of an originating cfDNA molecule;
a rate of strand-level errors where an expected microsatellite length of an antisense strand is one repeat unit longer than an expected microsatellite length of an originating cfDNA molecule;
a rate of strand-level errors where an expected microsatellite length of a sense strand is one repeat unit shorter than an expected microsatellite length of an originating cfDNA molecule; and/or
a rate of strand-level errors where an expected microsatellite length of an antisense strand is one repeat unit shorter than an expected microsatellite length of an originating cfDNA molecule.
9. The method of claim 5 , wherein γ comprises:
a rate of read-level errors where a microsatellite length observed within a sequencing read is one repeat unit longer than an expected microsatellite length for a strand of an originating cfDNA molecule; and/or
a rate of read-level errors where a microsatellite length observed within a sequencing read is one repeat unit shorter than an expected microsatellite length for a strand of an originating cfDNA molecule and/or wherein β comprises:
a rate of strand-level errors where an expected microsatellite length of a sense strand is one repeat unit longer than an expected microsatellite length of an originating cfDNA molecule;
a rate of strand-level errors where an expected microsatellite length of an antisense strand is one repeat unit longer than an expected microsatellite length of an originating cfDNA molecule;
a rate of strand-level errors where an expected microsatellite length of a sense strand is one repeat unit shorter than an expected microsatellite length of an originating cfDNA molecule; and/or
a rate of strand-level errors where an expected microsatellite length of an antisense strand is one repeat unit shorter than an expected microsatellite length of an originating cfDNA molecule.
10. The method of claim 1 , comprising determining the site specific trained threshold and/or the population trained threshold from sequence information from a population of microsatellite loci in one or more training DNA samples.
11. The method of claim 10 , wherein the one or more training DNA samples, comprises at least one of: (a) non-tumor cfDNA samples, and (b) DNA samples from one or more tumor types.
12. The method of claim 1 , wherein step (iv) comprises classifying the MSI status of the sample as MSI-high (MSI-H) if the microsatellite instability score exceeds the threshold in the at least one or more unstable microsatellite loci from the plurality of the microsatellite loci.
13. The method of claim 1 , wherein step (iv) comprises classifying the MSI status of the sample as MSI-high (MSI-H) if the number of unstable microsatellite loci comprises about 0.1%, about 1%, about 2%, about 3%, about 4%, about 5%, about 10%, about 15%, about 20%, or about 25% of the plurality of the microsatellite loci.
14. The method of claim 1 , wherein at least one of unstable microsatellite loci from the plurality of the microsatellite loci comprises at least 3 unstable microsatellite loci.
15. The method of claim 1 , wherein the antibody is pembrolizumab, nivolumab, ipilimumab, atezolizumab, avelumab, or durvalumab.
16. The method of claim 1 , wherein the wherein the immune checkpoint molecule comprises an inhibitory molecule that reduces a signal involved in the T cell response to antigen.
17. The method of claim 1 , wherein the immune checkpoint molecule comprises at least one of CTLA4, PD-1, PD-L1, PD-L2, CTLA4, CD80, CD86, lymphocyte activation gene 3 (LAG3), killer cell immunoglobulin like receptor (KIR), T cell membrane protein 3 (TIM3), galectin 9 (GAL9), or adenosine A2a receptor (A2aR).
18. The method of claim 1 , wherein the immune checkpoint molecule is a co-stimulatory molecule or a ligand of a co-stimulatory molecule comprising CD28, CD80, CD86, B7RP1, B7-H3, B7-H4, CD137L, OX40L, or CD70.
19. The method of claim 1 , wherein the human subject is afflicted with breast, colon or lung cancer.